[SciPy-User] Error when building scipy.0.9.0 - "f951: undefined symbol: mpfr_get_z_exp"

Bruce Southey bsouthey@gmail....
Mon Aug 29 08:42:44 CDT 2011


On 08/29/2011 06:20 AM, David Cournapeau wrote:
> On Mon, Aug 29, 2011 at 11:13 AM, Xiong Deng<dbigbear@gmail.com>  wrote:
>> Hi,
>>
>> I just find out that the gcc-4.7 is downloaded as a binary distri. I did not
>> compile gcc-4.7 myself...
> Then the binary is buggy or not adapted to your platform.
>
>> The gcc included on my system is gcc 3.4.5 and there seems no gfortran built
>> on it (However there is a g77 on it, which cause problems while building
>> numpy/scipy....).
> You should be able to build numpy and scipy with g77. You should not
> try mixing compiler versions unlesss you are willing to spend quite
> some time debugging subtle mismatches issues.
>
>> ..In addition, there are not mpc, mpfr, gmp on it with gcc
>> 3.4.5, so I need a new gcc with gfortran, mpc ,mpfr ,gmp, which is necessay
>> for numpy/scipy....
> The usual way to do this is to first build mpfr and gmp with whatever
> compiler you have (gcc 3.4.5 here), and then build the new version of
> gcc and gfortran. But again, you would be better just using the
> compilers you have on your machine.
>
> cheers,
>
> David
> _______________________________________________
> SciPy-User mailing list
> SciPy-User@scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-user
Assuming this is the same John or Johnny as before,
RHEL does provide Atlas as a package so you should get it from Red Hat 
package manager.

If you build things yourself, you must ensure that all previous versions 
have been removed from everywhere and that you are linking paths are to 
the correct locations.

Bruce
PS please use one name and one thread




More information about the SciPy-User mailing list